-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathroc_diagrams.py
146 lines (132 loc) · 5.83 KB
/
roc_diagrams.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
import rank_histograms as rh
import matplotlib.pyplot as plt
from matplotlib.ticker import FixedLocator
import xskillscore as xs
import numpy as np
def plot_roc(roc,model,event_type='split'):
false_positive=roc.sel(metric='false positive rate')
true_positive=roc.sel(metric='true positive rate')
#roc.to_dataset(dim='metric').plot.line(y='true positive rate', x='false positive rate')
#roc.to_dataset(dim='metric').plot.plot(y='true positive rate', x='false positive rate')
plt.plot(false_positive, true_positive,'o-')
plt.xlabel('false positive rate')
plt.ylabel('true positive rate')
area=roc.sel(metric='area under curve').values[0]
plt.plot([0, 1], [0, 1], 'k:')
#ax.text(0.5,1.08,'ROC curve '+event_type+' events '+model,fontsize=12,ha='center',transform=ax.transAxes)
#ax.text(0.5,1.02,'Area under curve: '+str(round(area,3)),fontsize=9,ha='center',transform=ax.transAxes)
plt.title('')
#estimate uncertainty with perfect model range
def freq_test(ds,threshold,lower=True):
nmem=len(ds['member'])
freq_stats=np.empty((nmem))
for mem in range(nmem):
observations=ds.isel(member=mem)
if lower:
roc_areas=roc=xs.roc(observations<threshold, (ds<threshold).mean('member'),
return_results='all_as_metric_dim')
if not lower:
roc_areas=roc=xs.roc(observations>threshold, (ds>threshold).mean('member'),
return_results='all_as_metric_dim')
area=roc.sel(metric='area under curve').values[0]
freq_stats[mem]=area
mean=np.mean(freq_stats[:])
perc5=np.percentile(freq_stats[:]-mean,5)
perc95=np.percentile(freq_stats[:]-mean,95)
return perc5,perc95
"""
obs=rh.get_dataset(rh.path_to_data+'ERA5_moments@10and50and100hPa_NHonly.nc')
obs_cl=obs['centroid_latitude'].loc['1979-01-01':'2014-12-28',1e3]
obs_ar=obs['aspect_ratio'].loc['1979-01-01':'2014-12-28',1e3]
area_displacements=[]
area_splits=[]
models=[]
for model in rh.file_name:
forecast=rh.get_dataset(rh.path_to_data+rh.file_name[model])
#define bin edges
#edges=np.arange(0,1.02,0.02)
if any(forecast['plev'].isin(1e3)):
#roc curve for displacement
forecast_cl=forecast['centroid_latitude'].loc[:,'1979-01-01':'2014-12-28',1e3]
roc=xs.roc(obs_cl<66, (forecast_cl<66).mean('member'),return_results='all_as_metric_dim') #thresholds as in Seviour et al. 2013
area=roc.sel(metric='area under curve').values[0]
#use rank frequency test to produce error bars
err_5,err_95=freq_test(forecast_cl,threshold=66,lower=True)
area_displacements.append('{model}&{area:.3f}&{perc5:.3f}&{perc95:.3f}'.format(
model=model,area=area,perc5=err_5,perc95=err_95))
#plot_roc(roc,model,event_type='displacement')
# roc.to_dataset(dim='metric').plot.scatter(y='true positive rate', x='false positive rate')
# plt.plot([0, 1], [0, 1], 'k:')
# plt.title('ROC curve displacement events '+model)
#plt.savefig('../figures/roc_diagrams/displacement_'+model+'.png')
#plt.close()
print('displacement',model)
#roc curve for split
forecast_ar=forecast['aspect_ratio'].loc[:,'1979-01-01':'2014-12-28',1e3]
roc=xs.roc(obs_ar>2.4, (forecast_ar>2.4).mean('member'),return_results='all_as_metric_dim')
area=roc.sel(metric='area under curve').values[0]
err_5,err_95=freq_test(forecast_ar,threshold=2.4,lower=False)
area_splits.append('{model}&{area:.3f}&{perc5:.3f}&{perc95:.3f}'.format(
model=model,area=area,perc5=err_5,perc95=err_95))
#plot_roc(roc,model)
#plt.savefig('../figures/roc_diagrams/split_'+model+'.png')
#plt.close()
models.append(model)
print('split',model)
"""
models=[]
area_displacements=[]
err_locs_disp=[]
with open('roc_displacement.txt','r') as f:
lines=f.readlines()
for line in lines:
line=line.split('&')
models.append(line[0])
area_displacements.append(float(line[1]))
err_locs_disp.append([abs(float(line[2])),abs(float(line[3]))])
print(err_locs_disp)
ind=np.arange(len(models))
width=0.5
fig, axs=plt.subplots(1,2,sharey=True,dpi=480,figsize=(8,4.5))
plt.suptitle('ROC area under curve')
for i in ind:
axs[0].bar(i,area_displacements[i],width,color='lightblue',edgecolor='black')
axs[0].errorbar(i,area_displacements[i],yerr=[[err_locs_disp[i][0]],[err_locs_disp[i][1]]],
ecolor='black',capsize=3,fmt='none')
axs[0].xaxis.set_major_locator(FixedLocator(ind))
axs[0].set_xticklabels(models,rotation=75)
axs[0].axhline(0.5,color='grey',linestyle='--')
axs[0].axvline(2.5,color='black',linestyle='-')
axs[0].set_ylabel('Area under curve')
axs[0].set_title('Displacement events')
#plt.savefig('../figures/areas_displacement.png')
#plt.close()
area_splits=[]
err_locs_splits=[]
with open('roc_split.txt','r') as f:
lines=f.readlines()
for line in lines:
line=line.split('&')
area_splits.append(float(line[1]))
err_locs_splits.append([abs(float(line[2])),abs(float(line[3]))])
for i in ind:
axs[1].bar(i,float(area_splits[i]),width,color='lightblue',edgecolor='black')
axs[1].errorbar(i,area_splits[i],yerr=[[err_locs_splits[i][0]],[err_locs_splits[i][1]]],
ecolor='black',capsize=3,fmt='none')
axs[1].xaxis.set_major_locator(FixedLocator(ind))
axs[1].set_xticklabels(models, rotation=75)
axs[1].axhline(0.5,color='grey',linestyle='--')
axs[1].axvline(2.5,color='black',linestyle='-')
#axs[1].set_ylabel('Area under curve')
axs[1].set_title('Split events')
fig.tight_layout()
plt.savefig('../figures/areas_displacement_split_new.png')
plt.close()
"""
with open('roc_displacement_new2.txt','w') as f:
for row in area_displacements:
f.write(row+'\n')
with open('roc_split_new2.txt','w') as f:
for row in area_splits:
f.write(row+'\n')
"""